i - cortical column functions ii - functional webs ling 411 – 12

Post on 18-Jan-2016

219 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

I - Cortical Column Functions

II - Functional Webs

Ling 411 – 12

Uniformity of cortical function

If cortical function is uniform across mammals and across different cortical areas, then the findings presented by Mountcastle can be extended to language

Claims:•Locally, all cortical processing is the same

•The apparent differences of function are consequences of differences in larger-scale connectivity

Conclusion (if the claim is supported):•Understanding language, even at higher

levels, is basically a perceptual process

Testing the claim

Claim:•The apparent differences of function are

consequences of differences in larger-scale connectivity

To test, we need to understand cortical function

That means we have to understand the function of the cortical column

Quote from Mountcastle

“[T]he effective unit of operation…is not the single neuron and its axon, but bundles or groups of cells and their axons with similar functional properties and anatomical connections.”

Vernon Mountcastle, Perceptual Neuroscience (1998), p. 192

Columns do not store symbols!

They only•Receive activation•Maintain activation• Inhibit competitors•Transmit activation

Important consequence:•We have linguistic information

represented in the cortex without the use of symbols

• It’s all in the connectivity Challenge:

•How?

Why the usual approach won’t work

Let us suppose that words are stored in some kind of symbolic form

What form? If written, there has to be..

•something in there that can read them•something in there that can write them•something in there that can move them

around, from one place to another•something in there to compare them with

forms entering the brain as it hears someone speaking – otherwise, how can an incoming word be recognized?

Why the usual approach won’t work (cont’d)

If not written, then represented in some other medium

Doesn’t solve the problem You still need whatever kind of sensory

detectors can sense the symbols in whatever medium you choose

Plus means of performing all those other operations

Compare imagery

Visual images•Little pictures?

• If so, what is in there to see them?

Auditory images•Little sounds vibrating in the brain?

• If so, what is in there to hear them?

There has to be another way!

There must be another way

Visual imagery (e.g. of your grandmother)•Reactivation of some of the same nodes and

connections that operate when actually seeing her

Auditory imagery (e.g. of a tune)•Reactivation of some of the same nodes and

connections that operate in actually hearing it

Another way, for language

A syllable•Activation of the nodes and connections

needed to recognize or produce it

A word•Activation of the nodes and connections

needed to recognize it

A syntactic construction•Activation of the nodes and connections

needed to recognize or produce it

The postulation of objects as something different from the terms of relationships is a superfluous axiom and consequently a metaphysical hypothesis from which linguistic science will have to be freed.

Louis Hjelmslev Prolegomena to a Theory of Language

(1943: 61)

Quotation

Columns do not store symbols!

They only•Receive activation•Maintain activation• Inhibit competitors•Transmit activation

Important consequence:•We have linguistic information

represented in the cortex without the use of symbols

• It’s all in the connectivity Challenge:

•How?

Columnar Functions: Integration and Broadcasting

Integration: A column is activated if it receives enough activation from • Other columns • Thalamus

Can be activated to varying degrees Can keep activation alive for a period of time Broadcasting: An activated column transmits

activation to other columns• Exitatory• Inhibitory

Learning : adjustment of connection strengths and thresholds

Integration and Broadcasting

Broadcasting•To multiple

locations

• In parallel

Integration

Integration and Broadcasting

Integration

Broadcasting

Wow, I got activated!

Now I’ll tell my friends!

What matters is not ‘what’ but ‘where’

What distinguishes one kind of information from another is what it is connected to

Lines and nodes are approximately the same all over

Hence, uniformity of cortical structure•Same kinds of columnar structure

•Same kinds of neurons

•Same kinds of connections

Different areas have different functions because of what they are connected to

Operations in relational networks

Activation moves along lines and through nodes• Integration

•Broadcasting

Connection strengths are variable•A connection becomes stronger with

repeated successful use

•A stronger connection can carry greater activation

What about the rest of language?

Words and their meanings Syntax and morphology Conceptual relationships

Sequence

In language, sequence is very important•Word order

•Order of phonological elements in syllables

•Etc.

Also important in many non-linguistic areas•Dancing

•Eating a meal

Can cortical columns handle sequences?

Lasting activation in minicolumn

Subcorticallocations

Connections to neighboring columns not shown

Cell Types

Pyramidal

Spiny Stellate

Inhibitory

Recurrent axon branches keep activation alive in the column –Until is is turned off by inhibitory cell

Notation for lasting activation

> Thick border for a node that stays active for a relatively long time > Thin border for a

node that stays active for a relatively short time

Recognizing items in sequence

This link stays active

a b

Node c is satisfied by activation from both a and b If satisfied it sends activation to output connections Node a keeps itself active for a whileSuppose that node b is activated after node a Then c will recognize the sequence ab

c

This node recognizes the sequence ab

Demisyllables in recognizing stops

Consider stop consonants, e.g. t, d At the time of closure

•For voiceless stops there is no sound to hear

•For voiced stops, very little sound

The stops are identified by transitions •To following vowel

•From preceding vowel

Demisyllables [di, de, da, du]

F1 and F2For [a]

It is unlikely that [d] is represented as a unit in perception

Recognizing a syllable and its demisyllables

dim

di- -im

Cardinal node for dim

Functional subweb for dim

Auditory features of [di-]Auditory features of [-im]

Just labels

Another syllable and its demisyllables

bil

bi- -il

Cardinal node for bill

Subweb for bill

Multiple connections of -il

bil hil kil

bi- -il

Bill hill mill kill etc.

One and the same /-il/ in all of them

Multiple connections of -il

bil hil kil

bi- -il

Bill hill mill kill etc.

Similarly for multiple connections of bi- bit, bib, bid, etc.

Multiple connections of -il

bil hil kil

bi- -il

Bill hill mill kill etc.

To lower level nodes in the subwebs, for phonological features

Syntactic Recognition – same principle

This link stays active

a b

Let node a represent Noun Phrases (Subject) and let b represent Predicates (Verb Phrases etc.)Then c represents Clauses: the sequence ab

c

This node recognizes the sequence ab

Syntactic Recognition: higher-level perception

This link stays active

a b

The whole process is one of recognition, just as at lower levels (e.g., phonological recognition)Same structures, different connections

c

This node recognizes the sequence ab

Conclusion: All of linguisticstructure is relational

The whole of linguistic structure is a connectionist system

Good thing, since that is exactly the kind of system that the cortex is built to represent and to operate with

Findings relating to columns(Mountcastle, Perceptual Neuroscience, 1998)

The column is the fundamental module of perceptual systems • probably also of motor systems

Perceptual functions are very highly localized• Each column has a very specific local function

This columnar structure is found in all mammals that have been investigated

The theory is confirmed by detailed studies of visual, auditory, and somatosensory perception in living cat and monkey brains

Operation of the Network

The linguistic system operates as distributed processing of multiple individual components – cortical columns

Columnar Functions • Integration: A column is activated if it receives

enough activation from other columns Can be activated to varying degrees Can keep activation alive for a period of time

• An activated column transmits activation to other columns Exitatory – contribution to higher level Inhibitory – dampens competition at same

level

Columns do not store symbols!

Review

Neuronal Structure and Function

(Pulverműller 2002, Chapter 2)

Neuronal Structure and Function:The Cortex as a Network

Pulvermüller (2002):•The brain is not like a computer“…any hardware computer configuration can

realize almost any computer program or piece of software.”

“… it may be that the neuronal structures themselves teach us about aspects of the computational processes that are laid down in these structures.”

Connectivity as key property

The cortex operates by means of connections

Grey matter•Cortical columns

•Horizontal connections among neighboring columns

White matter•Connections between distant

columns

Computers and Brains: Different Structures, Different Skills

Computers•Exact, literal

•Rapid calculation

•Rapid sorting

•Rapid searching

•Faultless memory

•Do what they are told

•Predictable

Brains•Flexible, fault tolerant

•Slow processing

•Association

• Intuition

•Adaptability, plasticity

•Self-driven activity

•Unpredictable

•Self-driven learning

What brains but not computers can do

Acquire information to varying degrees• “Entrenchment”

• How does it work? Variable connection strength Connections get stronger with repeated use

Perform at varying skill levels• Degrees of alertness, attentiveness

• Variation in reaction time

• Mechanisms: Global neurotransmitters Variation in blood flow Variation in available nutrients Presence or absence of fatigue Presence or absence of intoxication

Neuronal Structure and Function:Connectivity

White matter: it’s all connections•Far more voluminous than gray matter•Cortico-cortical connections

The fibers are axons of pyramidal neurons They are all excitatory

•White since the fibers are coated with myelin Myelin: glial cells

There are also grey matter connections•Unmyelinated•Local•Horizontal, through gray matter•Excitatory and inhibitory

Pyramidal neurons and their connections

Connecting fibers•Dendrites (input): length 2mm or less•Axons (output): length up to 10 cm

Synapses•Afferent synapses: up to 50,000

From distant and nearby sources•Distant – to apical dendrite•Local – to basal dendrites or cell body

•Efferent synapses: up to 50,000 On distant and nearby destinations

•Distant – main axon, through white matter•Local – collateral axons, through gray

matter

Proportion of pyramidal cells in the cortex

Abeles (1991: 52) says 70% Mountcastle says 70% - 80% (1998: 54)

•Based on information from Feldman (1984) Pulvermüller (2002: 13) says 85%

•Based on information from Braitenburg & Schüz (1998)

Some difference comes from how spiny stellate cells are counted•Pyramidal or not?

No discrete boundary between these categories

Connecting fibers of

pyramidal neurons

Apical dendrite

Basal dendrites

Axon

Interconnections of pyramidal neurons

Input from distant cells

Input from neighboring columns

Output to distant cells

Neuronal Structure and Function:Connectivity

Synapses of a typical pyramidal neuron:• Incoming (afferent) – 50,000 (5 x 104)

•Outgoing (efferent) – 50,000

Number of synapses in cortex:•28 billion neurons (Mountcastle’s estimate)

i.e., 28 x 109

Synapses in the cortex (do the math)•5 x 104 x 28 x 109 = 140 x 1013 = 1.4 x 1015

•Approximately 1,400,000,000,000,000

• i.e., over 1 quadrillion

Cortical connectivity properties

Probability of adjacent areas being connected: >70% (Pulvermüller p. 17)

•But if we count by minicolumns instead of cells the figure is probably higher, maybe close to 100%

Probability of distant areas being connected: 15-30% (p. 17)

•Distant areas: at least one intervening area

• In Macaque monkey, most areas have links to 10 or more other areas within same hemisphere

More cortical connectivity properties

Most areas are connected to homotopic area of opposite hemisphere

Most connections between areas are reciprocal

Primary areas not directly connected to one another, except for motor-somatosensory•Connections under central sulcus

Degrees of separationbetween cortical neurons or columns

For neurons of neighboring columns: 1 For distant neurons in same hemisphere

•Range: 1 to about 5 or 6 (estimate)

•Mostly 1, 2, or 3, especially if functionally closely related

•Average about 3 (estimate)

For opposite hemisphere•Add 1 to figures for same hemisphere

Probably, for any two columns anywhere in the cortex, whether functionally related or not, fewer than 6 degrees of separation

Neural processes for learning

Basic principle: when a connection is successfully used, it becomes stronger•Successfully used if another connection

to same node is simultaneously active Mechanisms of strengthening

•Biochemical changes at synapses•Growth of dendritic spines•Formation of new synapses

Weakening: when neurons fire independently of each other their mutual connections (if any) weaken

Neural processes for learning

A

B

C

If connections AC and BC are active at the same time, and if their joint activation is strong enough to activate C, they both get strengthened

(adapted from Hebb)

Synapses here get strengthened

Pulvermüller’s functional webs

For example, a web for the concept CAT Pulvermüller:

•A significant portion of the web’s neurons are active whenever the cat concept is being processed

•The function of the web depends on the intactness of its member neurons

• If neurons in the functional web are strongly linked, they should show similar response properties in neurophysiological experiments

(2002:26)

The neural basis of cognition

Earlier proposals (p. 23)

• Individual neurons (Barlow 1972) Individual neurons too noisy and unreliable Would require more information processing

capacity than one neuron has

• Mass activity and interference patterns in the entire cortex (Lashley 1950)

Better alternative:• Functional webs of neurons (Pulvermüller)

Even better• Functional webs of cortical columns

• (not mentioned by Pulvermüller)

Pulvermüller’s functional webs

A large set of neurons that

•Are strongly connected to each other

•Are distributed over a set of cortical areas

•Work together as a functional unit

•Are functionally interdependent so that each is necessary for the optimal functioning of the web (p.24)

Hypothesis I: Functional Webs

A word is represented as a functional web

Spread over a wide area of cortex• Includes perceptual information

Relating to the meaning

• As well as specifically conceptual information

• For nominal concepts, mainly in

• Angular gyrus

• (?) For some, middle temporal gyrus

• (?) For some, supramarginal gyrus

• As well as phonological information Temporal, parietal, frontal

Example: The meaning of dog

We know what a dog looks like•Visual information, in occipital lobe

We know what its bark sounds like•Auditory information, in temporal lobe

We know what its fur feels like•Somatosensory information, in parietal lobe

All of the above..•constitute perceptual information•are subwebs with many nodes each•have to be interconnected into a larger web•along with further web structure for

conceptual information

The Wernicke-Lichtheim concept node (1885)

Where?

The “C” Node

Not just in one place•Conceptual information for a single word is

widely distributed•Conceptual information is in different areas

for different kinds of concepts The second of these points and

probably also the first were already recognized by Wernicke

But.. •There may be a single “C” node anyway as

cardinal node of a distributed network

“C” node as cardinal node of a web

V

M

C

For example, FORK

Labels for Properties:C – ConceptualM – MotorT – TactileV - Visual

Each node in this diagramrepresents the cardinal node of a subweb of properties

T

Some connections of the “C” node for FORK

V

C

Each node in this diagramrepresents the cardinal node of a subweb of properties

For example,

Let’s zoom in on this one

M

T

Zooming in on the “V” Node..

FORK

Etc. etc.(many layers)

A network of visual featuresV

Add phonological recognition node

V

M

C

For example, FORK

Labels for Properties:C – ConceptualM – Motor P – Phonological imageT – TactileV – Visual

T

P

The phonological image of the spoken form [fork] (in Wernicke’s area)

Add node in primary auditory area

V

M

CT

P

PA

Primary Auditory: the cortical structures in the primary auditory cortex that are activated when the ears receive the vibrations of the spoken form [fork]

For example, FORK

Labels for Properties:C – ConceptualM – Motor P – Phonological imagePA – Primary AuditoryT – TactileV – Visual

Add node for phonological production

V

M

CT

P

PA

PP

For example, FORK

Labels for Properties:C – ConceptualM – Motor P – Phonological imagePA – Primary AuditoryPP – Phonological ProductionT – TactileV – Visual

Arcuate fasciculus

Articulatory structures (in Broca’s area) that control articulation of the spoken form [fork]

Some of the cortical structure relating to fork

V

M CT

P

PA

PP

Functional web of a simple lexeme: fork

V

MC

T

P

PA

PP

Phonological form

Meaning

Link betw form and meaning

Part of the functional web for FORK(showing cardinal nodes only)

V

MC

T

P

PA

PP

Each node shown here is the cardinal node of a subweb

For example, the cardinal node of the visual subweb

An activated functional web(with two subwebs partly shown)

V

PRPA

M

C

PP

T

Visual features

C – Cardinal concept nodeM – MemoriesPA – Primary auditoryPP – Phonological productionPR – Phonological recognitionT – TactileV – Visual

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Ignition of a functional web from visual input

V

PR

PA

M

C

Art

T

Speaking as a response to ignition of a web

V

PR

PA

M

C

Art

T

Speaking as a response to ignition of a web

V

PR

PA

M

C

Art

T

Speaking as a response to ignition of a web

V

PR

PA

M

C

Art

T

From here (via subcortical structures) to the muscles that control the organs of articulation

An MEG study from Max Planck Institute

Levelt, Praamstra, Meyer, Helenius & Salmelin, J.Cog.Neuroscience 1998

Pulvermüller’s line of reasoning

1. “If neurons in the functional web are strongly linked, they should show similar response properties in neurophysiological experiments.

2. “If the neurons of the functional web are necessary for the optimal processing of the represented entity, lesion of a significant portion of the network neurons must impair the processing of this entity. This should be largely independent of where in the network the lesion occurs.

3. “Therefore, if the functional web is distributed over distant cortical areas, for instance, certain frontal and temporal areas, neurons in both areas should (i) share specific response features and (ii) show these response features only if the respective other area is intact.”

(2002: 26, see also 27)

end

top related